Spectrum and Signal Analysis…Pulsed RF

Application Note 150-2

 Contents

Chapter 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
The basic pulse spectrum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
Chapter 2. “Line” Spectrum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
General rules and explanation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
Pulse desensitization αL . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Transition to the pulse response . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
Chapter 3. “Pulse” Spectrum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
General rules and explanation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
Why use a "pulse" spectrum display . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
Peak pulse response -pulse desensitization αp . . . . . . . . . . . . . . . . . . . . 15
Very short RF pulses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
Chapter 4. Summary of pulsed spectra characteristics . . . . . . . . . . . . . . . . . . . . . . . 18
Common pulse spectra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
Appendix A. Table of important transforms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
Explanation of the table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
Properties of transforms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
Appendix B. IF Amplifier Response and Distortion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
IF amplifier response . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
Distortion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

2
Chapter 1 – Introduction
The basic pulse spectrum
The spectrum analyzer was originally designed to look at the wave. The fundamental frequency determined the square wave
output of radar transmitters. A pulse radar signal is a train of RF
pulses with a constant repetition rate, constant pulse width and
shape, and constant amplitude. By looking at the characteristic
spectra, all-important properties of the pulsed signal such as
pulse width, occupied bandwidth, duty cycle, and peak and aver-
age power, can be measured easily and with high accuracy.
Perhaps an even more important application of the spectrum
analyzer is the detection of transmitter misfiring and frequency
pulling effects. This application note is intended as an aid for the
operation of spectrum and signal analyzers and the interpretation
of their displayed pulse spectra.
The formation of a square wave from a fundamental sine wave and
its odd harmonics is a good way to start an explanation of the
spectral display for nonsinusoidal waveforms. You will recall perhaps
at one time plotting a sine wave and its odd harmonics on a
sheet of graph paper, then adding up all the instantaneous values.
If there were enough harmonics plotted at their correct amplitudes
and phases, the resultant waveform began to approach a square
T Sum of fundamental,
E A
Time
Figure 1a. Periodic rectangular pulse train
rate, and the amplitudes of the harmonics varied inversely to their
number.
A rectangular pulse is merely an extension of this principle, and by
changing the relative amplitudes and phases of harmonics, both odd
and even, we can plot an infinite number of wave shapes. The
spectrum analyzer effectively separates waveforms and presents
the fundamental and each harmonic contained in the waveform.
Consider a perfect rectangular pulse train as shown in Figure 1a,
perfect in the respect that rise time is zero and there is no overshoot
or other aberrations. This pulse is shown in the time domain and
we wish to examine its spectrum so it must be broken down into
its individual frequency components. Figure lb superimposes the
fundamental and its second harmonic plus a constant voltage
to show how the pulse begins to take shape as more harmonics
are plotted. If an infinite number of harmonies were plotted, the
resulting pulse would be perfectly rectangular. A spectral plot of
this would be as shown in Figure 2.
2nd harmonic and
Average Fundamental average value
value of
wave 2nd harmonic
Figure 1b. Addition of a fundamental cosine wave and its
harmonics to form rectangular pulses
3
 A
T Spectral lines PRF

f
1
T -4 -3 -2 -1 1 2 3 4
1 2 3 4
DC ∞ fc
Frequency, f

Figure 2. Spectrum of a perfectly rectangular pulse. Amplitudes and Figure 3. Resultant spectrum of a carrier amplitude modulated with a
phases of an infinite number of harmonics are plotted, resulting in rectangular pulse
smooth envelope as shown.

The envelope of this plot follows a funtion of the basic form: pulse repetition frequency with its harmonics. Thus, the lines will
be spaced in frequency by whatever the pulse repetition frequency
sin( x ) might happen to be. The spectral line frequencies may be
y=
x
F=
L
FC ± nPRF
There is one major point that must be made clear before going
into the spectrum analyzer display further. We have been talking expressed as:
about a square wave and a pulse without any relation to a carrier
or modulation. With this background we now apply the pulse where Fc = Carrier frequency
waveform as amplitude modulation to an RF carrier. This produces
sums and differences of the carrier and all of the harmonic PRF = Pulse repetition frequency
components contained in the modulating pulse. n = 0, 1, 2, 3 . . . . . .
The “mainlobe” in the center and the “sidelobes” are shown as
We know from single tone AM how the sidebands are produced groups of spectral lines extending above and below the baseline.
above and below the carrier frequency. The idea is the same for For perfectly rectangular pulses and other functions whose
a pulse, except that the pulse is made up of many tones, thereby derivatives are discontinuous at some point, the number of
producing multiple sidebands that are commonly referred to as sidelobes is infinite.
spectral lines on the spectrum analyzer display. In fact, there
will be twice as many sidebands or spectral lines as there are The mainlobe contains the carrier frequency represented by the
harmonics contained in the modulating pulse. longest spectral line in the center. Amplitude of the spectral lines
forming the lobes varies as a function of frequency according to the
Figure 3 shows the spectral plot resulting from rectangular expression for a perfectly rectangular pulse.
amplitude pulse modulation of a carrier. The individual lines
represent the modulation product of the carrier and the modulating τ
sin(ω )
2
τ
ω
2

4
Thus, for a given carrier frequency the points where these lines
go through zero amplitude are determined by the modulating
pulse width only. As pulse width becomes shorter, minima of the
envelope become further removed in frequency from the carrier, and
the lobes become wider. The sidelobe widths in frequency are
related to the modulating pulse width by the expression f = 1/Τ
Because the mainlobe contains the origin of the spectrum (the
carrier frequency), the upper and lower sidebands extending from
this point form a mainlobe 2/Τ wide. Remember, however, that
the total number of sidelobes remains constant so long as the
pulse quality, or shape, is unchanged and only its repetition rate is
4a. Narrow pulse width causes wide spectrum lobes,
high PRF results in low spectral line density.
Figure 4c. PRF lower than 4a results in higher spectral
density. Lobe width is same as 4a since pulse widths but lobe widths are reduced by wider pulse.
are identical.
varied. Figure 4 compares the spectral plots for two pulse lengths,
each at two repetition rates with carrier frequency held constant.
Notice in the drawings how the spectral lines extend below the
baseline as well as above. This corresponds to the harmonics in
the modulating pulse, having a phase relationship of 180° with
respect to the fundamental of the modulating waveform. Since
the spectrum analyzer can only detect amplitudes and not phase,
it will invert the negative-going lines and display all amplitudes
above the baseline.
Figure 4b. Wider pulse than 4a causes narrower lobes, but
line density remains constant since PRF is unchanged.
Figure 4d. Spectral density and PRF unchanged from 4c,
5
 Because a pulsed RF signal has unique properties we have to be
careful to interpret the display on a spectrum analyzer correctly.
The response that a spectrum analyzer (or any swept receiver)
can have to a periodically pulsed RF signal can be of two kinds,
resulting in displays that are similar but of completely different
significance. One response is called a "line spectrum" and the
other is called a "pulse spectrum." We must keep in mind that
these are both responses to the same periodically pulsed RF
input signal, and the "line" and "pulse" spectrum refer solely to
the response or display on the spectrum analyzer.
We will discuss both types of response to a signal with the basic
appearance as shown in Figure 5 with the aid of pictures, and
then summarize all formulas and rules for proper operation of
the analyzer.

eff
Ep

t 1 = PRF
T

t=o p(t) cos ωot

t p(t) dt
eff = Width of Rectangular Pulse of same height and area as pulse applied to analyzer = fo Ep

Figure 5. Basic RF pulse

6
Chapter 2. “Line” Spectrum

General rules and explanation

A “line” spectrum occurs when the spectrum analyzer's 3 dB We will now look at the display images on page 9 of a pulsed
bandwidth RBW is narrow compared to the frequency spacing of RF signal to see how different sweep time, span and resolution
the input signal components. Because the individual spectral bandwidth of the spectrum analyzer influence the appearance of
components are spaced at the pulse repetition frequency (PRF) of the signal on the display.
the periodically pulsed RF, we can say:
A carrier signal with a CW amplitude of –30 dBm (Figure 7) is
RBW < PRF modulated by a pulse train with a PRF of 1 kHz and an effective
pulse width τeff of 0.1 ms (Figure 8). In Figure 9 we see the resulting
In this case all individual frequency components can be resolved.
pulse spectrum in a linear display. The RBW is 100 Hz, one-tenth
Only one is within the bandwidth at a time as shown in Figure 6.
of the PRF.
The display is a frequency domain display of the actual Fourier
The logarithmic display (Figure 10) allows a much better evalua-
components of the input signal. Each component behaves as a
tion of the signal spectrum, because the lower amplitudes of the
CW signal. The display has the normal true frequency domain
higher order sidelobes can now be easily measured.
characteristics:
Each Fourier component is resolved and the line spacing is measured
1. The spacing between lines on the display will NOT change
as 1 kHz, which is the PRF. We can also see that the spacing of
when the analyzer sweep time is changed.
the sidelobe minima is 10 kHz, according to the relation
2. The amplitude of each line will not change when the band 1 1
RBW is changed as long as RBW remains considerably = = 10 kHz
τ eff 0 .1 ms
below the PRF.

B < 1 or B < PRF

T

Figure 6. IF bandwidths smaller than PRF

7
 We thus can count ten spectral lines in each sidelobe or twenty
lines plus the carrier line in the mainlobe, according to the duty
cycle of the pulsed signal.
τ eff / T or τ eff PRF = 0 .1
(The fact that the amplitude of the spectral lines on the lobe called pulse desensitization.
minima reach zero for each integer ratio of τ eff / T can be
used to adjust the duty cycle very accurately.)
The display in Figure 10 does not change for different scan times,
unless we select a scan time too short for the given scan width
and bandwidth.
Most Agilent spectrum and signal analyzers have built-in logic
with a warning that enables us to avoid any wrong combination
of these control settings.
For spectrum analyzers without this feature we have to satisfy an
additional equation to avoid display errors:
Span[Hz / div]
< (RBW [Hz])2
SweepTime[s / div]
(See Appendix B)
In Figure 11 the RBW of the analyzer has been changed to 300 Hz.
Although the resolution of the spectral lines is reduced (minimal)
we still have a true Fourier line spectrum display. From this expe-
rience we can derive a rule of thumb for the analyzer's bandwidth
to obtain a line spectrum:
RBW < 0 .3 PRF (preferably RBW < 0 .1 PRF )
This rule is valid for the shape factors (4 to 4.5:1) of the IF
filters used in modern digital-IF spectrum analyzers.
In Figure 12 we have changed the span from 100 kHz (10 kHz/div)
=
to 50 kHz (5 kHz/div). We see that the spectrum envelope and
the line spacing have changed, but the number of lines in each
lobe remains constant. Written as a ratio in dB:
/ T0.1 ms to
In Figure 13 the pulse width has been altered from τ eff =
τ
/ T
eff = 0.05 ms. Comparing with Figure 10 (same control settings
on the spectrum analyzer), we find three differences:
1. The sidelobe minima are spaced by 20 kHz.
2. The number of lines in each sidelobe is 20. (The line
spacing is still 1 kHz since we did not change the PRF.)
3. The amplitude of the spectrum envelope is 6 dB lower.
8
The last point reveals a very important fact which has not been
mentioned yet, but can easily be seen in the display of the
calibrated logarithmic displays on page 9: The amplitude of the
carrier component (highest amplitude in the spectrum envelope)
of a pulse modulated signal is considerably lower than the CW
amplitude of the unmodulated carrier. This effect is commonly
Pulse desensitization αL
The expression “pulse desensitization” is quite misleading since
the sensitivity of the spectrum analyzer is not reduced by a pulse
modulated signal. The apparent reduction in peak amplitude can
be explained in the following manner: pulsing a CW carrier
results in its power being distributed over a number of spectral
components (carrier and sidebands). Each of these spectral com-
ponents then contains only a fraction of the total power.
In Figure 10, where we have a duty cycle τ eff / T of 0.1, we mea-
sure a display amplitude which has a difference of –20 dB com-
pared to the CW amplitude of the carrier. In Figure 13, with a duty
cycle of 0.05, we measure –26 dB. This leads to the equation for
the line spectrum pulse desensitization factor αL:
α L [dB] = (
20log10 τ eff / T )
(
= 20log10 τ eff PRF )
This relation is only valid for a true Fourier line spectrum
(RBW < 0.3 PRF). We can see that αL is only dependent on the
duty cycle τ eff / T of the pulsed signal.
The average power Pavg of the signal is also dependent on the
duty cycle:
Pavg P=
peak τ eff / T or Pavg Ppeakτ eff PRF
Pavg
dB = 10log10 (τ eff PRF )
Ppeak  
Line spectra of a pulsed modulated 1 GHz carrier

Figure 7. CW signal at 1 GHz with an Figure 8. Time domain display of the 1 GHz, Figure 9. Line spectrum of the pulsed 1 GHz
Amplitude of –30 dBm, span of 100 kHz, pulse width of 0.1 ms, period of 1.0 ms and a with 100 kHz span and linear amplitude
RBW of 100 Hz and reference of –20 dBm 5 ms sweep time display

Figure 10. Log display of signal in figure 9 Figure 11. Same signal as figure 9 with Figure 12. Same signal as figure 11 with span
bandwidth increased to 300 Hz decreased to 50 kHz and bandwidth decreased
to 100 Hz

Pavg
αL [dB] [dB]
Ppeak
– 80 – 40

– 70

– 60 – 30

– 50

– 40 – 20

– 30

Figure 13. The pulse width is changed to – 20 – 10

0.05 ms, pulse period of 1 ms and span
of 100 kHz – 10

0 0
10 – 4 10 – 3 10 – 2 10 – 1 1
or x PRF
T

Figure 14. Pulse desensitization αL (line spectrum)

Note: Images were captured using an Agilent PSA spectrum analyzer.

9
 Transition to the "pulse" response
We read from the diagram that for a duty cycle of 0.1 we will get If we increase the RBW in our example to 1 kHz, we get the dis-
a display desensitization of –20 dB, and for a ratio of 0.05 we get play shown in Figure 15. We notice that the analyzer has lost the
–26 dB as shown in Figure 10 and Figure 13. The diagram also ability to resolve the spectral lines because RBW = PRF. The
shows that the desensitization factor αL becomes very large for lines now displayed are generated in the time domain by the single
low duty cycles. In this case, the sensitivity of the analyzer and pulses of the signal. We also see that the displayed amplitude of
the maximum signal level at the broadband front end mixer the spectrum envelope has increased. This is due to the fact that
become important factors. We shall describe the necessary the IF filter is now sampling a broader part of the spectrum at a
considerations for these analyzer properties in the next chapter time, thus collecting the power of several spectral lines.
about the more important “pulse” spectrum display.
A pulse repetition rate equal to the resolution bandwidth is the
demarcation line between a true Fourier-series spectrum, where
each line is a response representing the energy contained in that
harmonic, and a “pulse” or Fourier-transform response.

Figure 15. The bandwidth was changed to equal the pulse

repetition frequency and the pulse width was changed to 0.1 ms.

10
Chapter 3 – “Pulse” Spectrum

General rules and explanation

A “pulse” spectrum occurs when the bandwidth RBW of the
spectrum analyzer is equal to/or greater than the PRF. The spectrum
analyzer in this case cannot resolve the actual individual Fourier
frequency domain components, since several lines are within its
bandwidth. However, if the bandwidth is narrow compared to the
spectrum envelope, then the envelope can be resolved (Figure 16).
The resultant display is not a true frequency domain display, but
a combination of time and frequency display. It is a time domain
display of the pulse lines because each line is displayed when a
pulse occurs, regardless of the frequency within the pulse spec-
trum to which the analyzer is tuned at that moment. It is a fre-
quency domain display of the spectrum envelope. The display has
three distinguishing characteristics:
1. The spacing between the pulse lines and their number will
change when the sweep time of the analyzer is changed. The
lines are spaced in real time by 1/PRF. The shape of the
spectrum envelope will not change with the scan time.
2. The spacing between the lines will not change when the span
is changed. The spectrum envelope will change horizontally as
we would expect.
3. The amplitude of the display envelope will increase linearly as
the bandwidth RBW is increased. This means an amplitude
increase of 6 dB for doubling RBW.
This is true as long as RBW does not exceed τ0.2 . When the
eff
bandwidth equals τ1 (or 1/2 of the mainlobe width), the display
eff
amplitude is practically the peak amplitude of the signal. At this
point the IF filter covers nearly all significant spectral components.
But then we have lost the ability to resolve the spectrum
envelope.

RBW > 1 or RBW > PRF

T

Figure 16. IF bandwidth greater than PRF

11
 We show these characteristics in the following pictures:
In Figure 17 we modulate the –30 dBm CW carrier by a pulse
1 1. For a sufficient resolution of the spectrum envelope the
train with a PRF of 100 Hz and τ eff = 10 kHz = 100 µ s .
0 .1
The analyzer's IF bandwidth is RBW = 1 kHz; i.e.,
τ eff
We can see the spectrum envelope with the mainlobe and side-
lobes and the minima in between. The lines that form the enve-
lope are not spectral lines but pulse lines in the time domain.
We can verify this by changing the scan time (Figure 18). If we
reduce the scan time further, we lose the information about the
shape of the spectrum envelope; i.e., the frequency domain infor-
mation. But we now can easily measure the PRF in the time
domain (Figure 19 and Figure 20).
In Figure 21 we changed the span to 5 kHz/div. The span is the
same as in Figure 18. We can see that the spectrum envelope
changed (frequency domain), but the line spacing remains con-
stant (time domain).
In Figure 22 we use an RBW of 300 Hz. We can measure an
amplitude decrease of approximately 10 dB compared to Figure
17, which shows the linear relationship between RBW and dis-
play amplitude. We also can see that the minima are better
resolved than in Figure 17. In Figure 23 the RBW is increased to
3 kHz. The display amplitude increase compared to Figure 22 is
not 20 dB but only 18 dB. We lost the linear relationship between
bandwidth and display amplitude because RBW is greater
0 .2
than in this case.
τ eff
Also the resolution of the sidelobes is lost to a great extent.
1
If we increase RBW to 10 kHz (which is equal to
τ eff
we get a display with an amplitude practically equal to the peak
amplitude of the pulsed signal (Figure 24).
12
Some additional rules of thumb are of importance:
bandwidth should be less than 5% of the mainlobe width
0 .1
. RBW <
τ eff
or:
For higher resolution into the lobe minima (20 to 30 dB) we
0 .03
RBW <
τ eff
should use:
2. The system must respond to each pulse independently.
The effects of one pulse must decay out before the next
pulse occurs. The IF amplifier decay time constant is
approximately 0.3/RBW. A decay of the pulse effect down
to 1% (–40 dB) requires five time constants. This leads
RBW > 1 .5PRF
to the rule:
However, we get less than 1 dB error if RBW = PRF, where
the baseline is only 20 to 25 dB below the spectrum
envelope (see Figure 15). The range between
RBW < 0.3PRF (line spectrum) and RBW > PRF (pulse
spectrum) shows properties of both response types and
should be avoided.
3. The number of pulse lines that form the spectrum
envelope display is determined by the PRF and the scan
10
scan time s/div  ≥
PRF [Hz]
), time. For a display with useful resolution, i.e., a sufficient
number of lines, the scan time should be selected to:
We then have more than 100 lines forming the spectrum enve-
lope, thus assuring that the mainlobe peak is displayed on each
scan (see Figure 18 and Figure 22). In the other displays where
1
RBW < we again have to consider a“pulse desensitization”
τ eff
factor because we compare a CW signal with a pulsed signal.
This factor will be extensively discussed later.
Pulsed RF signal in "pulsed" spectrum display
(All pictures show the same log ref of –20 dBm).

Figure 17. The PRF was set to 100 Hz (pulse peri- Figure 18. Same signal as figure 17 with the Figure 19. Same signal as figure 18 with the
od of 10 ms) pulse width of 0.1 ms, span of sweep time reduced to 1 sec sweep time future reduced to 200 ms
100 kHz, sweep time of 10 sec and RBW of 1 kHz

Figure 20. Same signal as figure 19 with the Figure 21. Same signal as figure 20 with the Figure 22. Same signal as figure 21 with the
RBW expanded to 300 kHz and sweep time 20 ms span of 50 kHz, RBW of 1 kHz and sweep time RBW reduced to 300 Hz, span increased to
of 1 sec 100 kHz, and sweep time increased to 2 seconds

Figure 23. The RBW is increased to 3 kHz Figure 24. The RBW is increased further to
10 kHz

13
 Why use a “pulse” spectrum display?
In many instances, it is neither possible nor desirable to make
a fine grain line-by-line analysis of a spectrum. A good example
of such a case is a train of short RF pulses at a low repetition
frequency as normally used in radar transmitters. Not only must
the IF bandwidth become inconveniently narrow, but often the
frequency modulation on the pulsed carrier is so excessive that
the resulting display is confusing.

In the "pulse" spectrum mode we can get all information we

need: the spectrum envelope and amplitude in the frequency
domain and the PRF in the time domain. We also have two
advantages over the “line” spectrum display:

1. We can use shorter scan times because of the greater bandwidth.

2. We can increase the display amplitude of the pulsed signal by

choosing a broader bandwidth. We know that the display ampli-
tude increases linearly with the bandwidth RBW. The noise level Figure 25. A 1 GHz carrier with –50 dBm amplitude, PRF of 400 Hz,
of the analyzer increases only proportional to √RBW. So we can 3 μsec pulse, RBW of 3 kHz, 5 MHz span, and a sweep time of 1
increase the signal-to-noise ratio proportional to √­RBW. The second is displayed. You will note that only the lower order
reason that the amplitude of the pulse increases proportional sidelobes can easily be measured.
to the RBW is that the signals are coherent and add directly. The
noise is not coherent and the power increases with the √­RBW.

This is opposite to the CW and “line” case where we have to use

narrower bandwidths to decrease the noise level, thus increasing
the signal-to-noise ratio. Figure 25 and Figure 26 show these
effects clearly.

From the preceding discussion about the “pulse” spectrum

response we can find another important fact: the spectrum/sig-
nal analyzer must provide independent controls for bandwidth,
span, and sweep time to optimize the display according to the
rules of thumb given for this type of response. Also the variable
persistence CRT offers a great advantage if we want to have a
flicker-free display of pulsed signals with low PRF.

Figure 26. The same display as figure 25 with the RBW increased
to 30 kHz. This increase resulted in a 20 dB increase in the
signal with only a 10 dB increase in noise. The higher order
sidelobes are now more easily measured.

14
Peak pulse response - pulse desensitization αp
In the "pulse" spectrum just described, the response of the
spectrum analyzer to each RF input pulse is in essence the pulse Bimp
response of the analyzer's IF amplifier.
E
The peak pulse response of Agilent spectrum and signal analyzers
has been established and is relatively independent of pulse shape
and pulse repetition frequency (for RBW > PRF). The expression
RBW3dB
relating the peak pulse response to a CW signal response is the
pulse desensitization factor αp.

This factor αp for the "pulse" response depends on different

physical conditions compared to αL in the "line" spectrum:

αp = ( )
20log10 τ eff Β imp [dB] f

In this equation we find a new expression: the effective impulse

bandwidth, Bimp. This can be visualized as the bandwidth of an Figure 27. Equivalent Bimp of Gaussian filter.
ideal, rectangularly shaped filter with a pulse response equivalent
to the actual filter with the 3 dB bandwidth B (Figure 27). Since There are several conditions which must be satisfied if this is to
the impulse bandwidth Bimp of the IF amplifier is not the same as be valid:
its 3 dB bandwidth B, a correction factor K has been introduced.
This factor K represents an empirical approach defining Bimp 1. The IF bandwidth-pulse width product must be less than
relative to B: two-tenths:
0 .2
Β imp RBW τ eff < 0 .2 or RBW <
K= τ eff
RBW

For 4-pole synchronously tuned filters as used in Agilent ESA 2. The normalized scan rate (NSR) of the analyzer must be
Series, 8560 family, and 859x IF sections, the value of K is ideally less than one:
1.6 with tolerance expectations of about ±10%, or ±0.8 dB. For ScanWidth [Hz]
=
the Agilent PSA Series, K is 1.479 with a tolerance of about ±2%
NSR <1
ScanTime s  × (RBW Hz )2
(±0.17 dB).

We can now write: 3.

The IF bandwidth must be greater than the PRF:
RBW > PRF
α p [dB] = ( )
20log τ eff K RBW ; K = 1 .48 or 1 .6
The conditions in 1 to 3 are automatically accomplished if the
Equations 5, 6, and 7 are satisfied.

4. The peak pulse amplitude at the broadband input mixer

of the analyzer must stay below the saturation point
(1 dB compression). The typical saturation point for Agilent
spectrum analyzers is between –10 dBm and +5 dBm:

Ppeak ≤ −10 dBm

15
 Figure 28 is a diagram showing the pulse desensitization αp in 1. We assume a pulsed signal with the following characteristics:
relation to IF bandwidth B and pulse width τeff . We see that the Ppeak = –30 dBm, τeff = 1 μs, PRF = 1 kHz. The noise level
PRF does not appear, since it is of no significance for the display of the analyzer is N = –100 dBm for 1 kHz bandwidth. We
amplitude as long as B > PRF. The shaded area between the find on the diagram for τeff = 1 μs an optimum bandwidth
of 100 kHz (→ RBW > PRF). We then can read a pulse
0 .03 0 .1
RBW = and RBW = represents the optimum bandwidth desensitization of αp≈ –16 dB. The displayed amplitude of
τ eff τ eff
the spectrum envelope will be ≈ –46 dBm. We also read
range for an analysis of a pulsed signal. There are also three from the crossing point of the line for N = –100 dBm and
dotted lines that show different noise levels of an analyzer for a the line for RBW = 100 kHz a resultant noise level of
fast determination of the dynamic range. –80 dBm. We thus get a usable display range (S/N ratio) of
only 34 dB. Although this range is sufficient in most cases
We will now take a few examples to show how the diagram for evaluation of the pulse spectrum, this example shows
is used: how important a spectrum analyzer with a low noise
level is.

αp [dB]
N [dB]
110 dBm noise level N of analyzer (RBW = 1 kHz)
–110
–100 dBm
–100
–90 dBm
–90 RBW > 1.7 PRF

–80
B
=
–70 1k
3k Hz
10 Hz
–60 kH
30 z
10 kH
–50 z
0k
30 Hz
–40 0k
1M Hz
–30 3 M Hz
Hz RBW = 0.03
τeff
–20
RBW = 0.1
τeff
–10

0
0.1 1 ns 10 100 1 µs 10 100 1 ms
τ eff
10 GHz 1 GHz 100 MHz 10 MHz 1 MHz 100 kHz 10 kHz
mainlobe width

Figure 28. Pulse desensitization αp (pulse spectrum)

16
 Very short RF pulses
2. Pulse power measurements: We see on the spectrum analyzer We know from the diagram for αp (Figure 28) that the desensitization
display the spectrum envelope of a pulsed signal with the of the analyzer display becomes very high for very short RF puls-
following characteristics: the display amplitude is –50 dBm, es, even with the widest resolution bandwidth. If we assume that
the mainlobe width is 10 MHz. The analyzer's bandwidth is we can provide the maximum usable input signal level of –10 dBm
300 kHz. What are the peak and the average powers of (which is normally possible when we measure in the proximity of
the signal? the radar transmitter to be investigated), we are then limited only
by the sensitivity of the analyzer. For a sufficient evaluation of a
The effective pulse duration τeff is calculated from the pulsed RF signal we should have a display range of at least 30 dB
lobe width or read from the diagram: above the noise level. Figure 29 is a diagram which shows the
2 maximum usable display range as a function of pulse width and
τ eff= = 0 .2 µs or 200 ns analyzer sensitivity for a maximum input level of –10 dBm and a
10 MHz
bandwidth of 300 kHz.*
In the diagram, we find a pulse desensitization of –21 dB
for τeff = 200 ns and RBW = 300 kHz. The peak power is We can easily see that for a pulse width of, for example, 1 ns, an
21 dB greater than the displayed amplitude, and we can analyzer must have a sensitivity of –110 dBm (specified for RBW
calculate the peak power to Ppeak = –29 dBm. = 300 kHz) or better to yield a usable display. It is not possible to
improve the signal-to-noise ratio with a low noise preamplifier,
To find the average power, we first have to measure the because we are already limited by the saturation level of the
PRF. This is done by reducing the scan time until we can input mixer. The new generation of Agilent spectrum analyzers
easily measure the pulse line spacing in time domain. offers exceptionally high sensitivities that allow measurements of
Assume we measured the line spacing to 1 ms which extremely short RF pulses.
equals a PRF of 1 kHz, we then can calculate the average
power Pavg = Ppeakτ eff PRF. *See page 15.

τ eff PRF = 2 × 10−7s × 103 Hz = 2 × 10−4

Using the diagram for αL, Figure 14 on page 9, we find a

Pavg 60
factor P of –37 dB. Thus, with the peak power Ppeak
peak Hz
50 0 0k
of –29 dBm and the factor of –37 dB, we can calculate the /3
Bm
average power Pavg = – 66 dBm. 20d Bm
–1 0d
display range [dB]

40
= 1
N –1 Bm
3.
We want to calculate the peak power of a signal displayed 0d
0 Bm
with an amplitude of –30 dBm and a mainlobe width of 30 –1
0d
100 MHz. The analyzer bandwidth is 300 kHz. The signal –9
2 20
=
has a pulse duration τ eff = 20 ns . Ppeak = –10 dBm
100 MHz
10 RBW = 300 kHz
We find a desensitization factor of – 41 dB.
0
This would yield a signal peak power of +11 dBm, far beyond 0.1 1 10 100
the saturation level of –10 dBm. Thus, the calculation is not τeff [ns]
valid. We have to insert at least 20 dB attenuation before
the input mixer. Figure 29. Display range vs. sensitivity

To check that the input signal level at the front end mixer is
below the saturation point, we have to observe that for a
10 dB step of the input attenuator the display amplitude
must also change by exactly 10 dB.

17
 Chapter 4
Summary of Pulse Spectra Characteristics

τeff
Ep

t=o p(t) cos ω ot

Type of response "Line" spectrum (Fourier series) "Pulse" spectrum (Fourier transform)
Requirements for each
Type of spectrum:
0 .1
Bandwidth RBW < 0.3PRF < RBW < 1 .7PRF
τ eff
Fs 10
Sweep time Ts > Ts >
RBW 2 PRF

Peak input power Ppeak ≤ −10 dBm Ppeak ≤ −10 dBm

τ 
Desensitization factor α L =20log10  eff
 T


αp = (
20log10 τ eff K RBW )
 

Amplitude of spectrum display at ω = ωo = ' eff

A E=
τ
E'pτ eff PRF A = E'pτ eff K RBW
p
T

Type of display used Fourier or spectral lines Pulse repetition rate lines
τ eff
for duty cycle > 0.05 < 0.05
T

Number of lines/Division Changes with scan width not scan time Changes with scan time not scan width

E’p = response on CRT due to CW signal Ep cos ωot RBW = IF bandwidth (3 dB)

Ts = sweep time K = constant of IF amplifier (K ≈ 1.479 or 1.617)

Fs = span τeff = width of rectangular pulse of same height and area as pulse
t p(t) dt
applied to analyzer = ∫
PRF = 1 = Pulse repetition frequency in Hz o E’p
T

18
 Common pulse spectra
Figure 30 shows some examples of typical spectrum displays for
pulse signals with different pulse shapes and with the presence
of AM and FM. An extensive mathematical treatment of different
pulse forms and their spectra can be found in Appendix A.
PULSE SPECTRUM IN THE PRESENCE OF FM: A symmetrical
pulse with linear coherent FM will produce a symmetrical spec-
trum with increased sidelobe amplitude and minima not reaching
zero, (c), (d).

The ideal rectangular RF pulse free of FM will produce a symmetri- If incidental FM (FM due to amplitude modulation) or coherent
cal pulse spectrum as shown in (a). When the pulse is changed FM is introduced together with an unsymmetrical pulse, an
to a triangular shape, the spectrum remains symmetrical with unsymmetrical pulse spectrum with the minima not reaching zero
decreased amplitude of the sidelobes (b). The pulse spectrum will be produced, (e), (f). This is also true for a symmetrical pulse
will remain symmetrical even if the pulse shape is distorted or with nonlinear coherent FM.
unsymmetrical.

1.0 Ampl.
1 µs ∆ F = 2 MHz/s
A f Freq.
spectrum amplitude

0.8 Freq.
Ampl. Fo
F
0.6 t t O
0.4
1.0 µs 1.0 µs
0.2
∆F=2

–4 –3 –2 –1 0 1 2 3 4
(b) (c)
(a)

Spectrum of rectangular pulse without AM Triangular pulse spectrum without FM Spectrum of rectangular pulse with linear
or FM occurring during pulse. Shape is that during pulse. Effective pulse width is FM resulting in increased sidelobe amplitude
t shorter than (a) causing minimas to occur and minimas not reaching zero.
sin(ω )
2 at wider intervals of frequency.
of t function.
ω
2

∆ F = 4 MHz/s .73A
∆ F = 2 MHz/s
Fo .73A Fo
∆ F = 2 MHz/s A
O Fo O
A
1 µs O

∆f=4 ∆f=2 ∆f=6

(d) (e) (f)

Same pulse spectrum as (c) with more Effect of linear AM and FM during pulse. More severe case of FM and AM occurring
severe FM. Note loss in symmetry due to pulse during pulse.
amplitude slope.

Figure 30. Common pulse spectra

19
 Figure 31: This is an example of a barker code radar showing the Figure 32: This is an example of a linear FM chirp radar showing
spectrum, IQ, pulse envelope and phase Barker. spectrum, pulse envelope, phase versus time and frequency versus time.

Figure 33: This spectrum is a result of AM and FM modulating the pulse. Figure 34: The transmitter is readjusted to reduce AM and FM.

20
Appendix A. Table of Important Transforms
Explanation of the table
The time functions and corresponding frequency functions in this
table are related by the following expressions:
∞ a simple RC cutoff. A one-coulomb impulse (pair 1S) applied to
F (ω ) = ∫ f (t )e
−iωt dt
(Direct transform)
−∞
∞
1
f (t ) = ∫ F (ω )e−iωt dω (Inverse transform)
2π −∞
The 1/2  multiplier in the inverse transform arises merely
because the integration is written with respect to ω, rather than
cyclic frequency. Otherwise the expressions are identical except
for the difference of sign in the exponent. As a result, functions
and their transforms can be interchanged with only slight modifi-
cation. Thus, if f(ω) is the direct transform of f(t), it is also true
that 2 f(– ω) is the direct transform of f (t). For example, the
spectrum of a sinx x pulse is rectangular (pair 6, page 22) while the
spectrum of a rectangular pulse is of the form sin x (pair 7, page 23).
x
Likewise pair 1S is the counterpart of the well-known fact that
the spectrum of a constant (dc) is a spike at zero frequency.
The frequency functions in the table are in many cases listed themselves are added, multiplied, or convolved. These relations
both as functions of ω and also of p. This is done merely for state mathematically many of the operations encountered in
convenience. F(p) in all cases is found by substituting p for iω in
F(ω). (Not simply p for ω as the notation would ordinarily indicate.
That is, in the usual mathematical convention one would write
F (ω ) = F ( −ip ) = G( p) where the change in letter indicates the
resulting change in functional form. The notation used above has
grown through usage and causes no confusion, once understood.)
Thus, in the p-notation
∞ i∞
1 a symmetry exists in these properties. Thus, delaying a function
F ( p) = ∫ f (t )e− pt dt f (t ) = F ( p)e pt dp
−∞ 2πi −∫i∞
The latter integral is conveniently evaluated as a contour integral
in the p-plane, letting p assume complex values.
The frequency functions have been plotted on linear amplitude
and frequency scales, and where convenient, also on logarithmic
scales. The latter scales often bring out characteristics not evi-
dent in the linear plot. Thus, many of the spectra are asymptotic
to first or second degree hyperbolas on a linear plot. On a log
plot these asymptotes become straight lines of slope –1 or –2
(i.e., –6 or –12 dB/octave).
The time functions in the table have all been normalized to con-
venient peak amplitudes, areas or slopes. For any other amplitude,
multiply both sides by the appropriate factor. Thus, the spectrum
of a rectangular pulse 10 volts in amplitude and 2 seconds long is
sin(ω )
(from pair 7, page 23) 20 20 volt-seconds.
ω F(p) = p 1+ α –
Again, upon multiplication by a constant having appropriate
dimensions, the frequency functions become filter transmissions.
Thus, if pair 1 is multiplied by α, the frequency function represents
this filter would produce an output (impulse response) with the
α
spectrum p + α × 1 coulomb, representing the time function
αe –ατ coulombs (which has the dimensions of amperes). Or a
1 volt step function (pair 2S) would produce the output spectrum
α 1
volts, which represents the time function (1 − e− at )
p+α p
volts (pair 4S).
The entries 1S through 6S in the table (page 24) are singular functions
for which the transforms as defined above exist only as limits. For
example, 1S may be thought of as the limit of pair 7 (multiplied
by 1 ) as τ → 0.
τ
Properties of transforms
There are a number of important relations which describe what
happens to the transforms of functions when the functions
communications systems: operations such as linear amplification,
mixing, modulation, filtering, and sampling. These relations are all
readily deducible from the defining equations above; but for ready
reference some of the more important ones are listed in the prop-
erties of transforms on the last page of this appendix.
Again, because of the similarity of the direct and inverse transforms,
multiplies its spectrum by a complex exponential; while multiplying
the function by a complex exponential delays its spectrum.
Multiplying any two functions is equivalent to convolving their
spectra; multiplying their spectra is equivalent to convolving the
functions.
Many of the pairs listed in the table of transforms can be obtained
from others by using one or more of the rules of manipulation listed
in the properties of transforms. For example, the time function in
pair 8, page 23 is 1 times the convolution of that in pair 7 with itself.
τ
The spectrum should therefore be τ1 times the product of that in
pair 7 with itself, as it indeed is. Further, by using these properties,
many pairs not in the table can be obtained from those given. For
example, the spectrum of f ( t )= (1 − α t ) e− at is (by the
addition property)
p
α =
(p + α)2 (p + α)2
21
 Table of Important Transforms

22
Table of Important Transforms

23
 Table of Important Transforms

24
Properties of Transforms

Time operation Frequency operation Significance

Linear addition Linear addition Linearity and superposition apply in both domains. The spectrum of a linear sum of
functions is the same linear sum of their spectra (if spectra are complex, usual rules of
af ( t ) + bg ( t ) aF (ω ) + bG (ω )
addition of complex quantities apply). Further, any function may be regarded as a sum of
component parts and the spectrum is the sum of the component spectra.

Scale change Inverse scale change Time-bandwidth invariance. Compressing a time function expands its spectrum in
f (kt ) 1 ω frequency and reduces it in amplitude by the same factor. The amplitude reduces because
F( ) less energy is spread over a greater bandwidth. For the same energy pulse as for k = 1 ,
k k
multiply both functions by k . The case where k = −1 reverses the function in time.
This merely interchanges positive and negative frequencies; so for real time functions,
reverses the phase.
1
Even and odd Even and odd Any real function f(t ) may be separated into an even part f ( t ) + f (−t ) and an odd part
2
partition partition 1
f ( t ) − f (−t ) . The transform of the even part is 1 F (ω ) + F (−ω ) which is purely real and
2 2 
1 1
f (t ) ± f (−t ) F (ω) ± F (−ω) 1
2 2 involves only even powers of ω. The transform of the odd part is F (ω ) − F (−ω) which is
2
purely imaginary and involves only odd powers of ω. Note: for f (t ) real, F ( −ω ) =
F* (ω ) .

Delay Linear added phase Delaying a function by a time to multiplies its spectrum by e−iωt0 , thus adding a delay of
f (t − t0 ) −iωt0 θ = −ωt0 to the original phase. Conversely a linear phase filter produces a delay
e F (ω )
of − dθ =
t0 .
dω
Multiplying a time function by eiωt0 f“delays”
(t ) its spectrum, i.e., shifts it to center about
Complex modulation Shift of spectrum
ωo rather than zero frequency. Ordinary real modulation — by cos(ω0t ) say — produces
iωt0 F (ω − ω0 )
e f (t )
the time function
2
(
1 iω0t
e )
+ e 0 f (t ) with the spectrum 1 F ω − ω0 + F (ω + ω0 ) .
−iω t
2
( ) 

Convolution The spectrum of the convolution of two time functions is the product of their spectra. In
Multiplication
convolution one of the two functions to be convolved is reversed left-to-right and displaced.
(filtering)
∞
The integral of the product is then evaluated and is a new function of the displacement.
F (ω ) G(ω ) Convolution occurs whenever a signal is obtained which is proportional to the integral of
∫ f (τ )g(t − τ )dτ
−∞ the product of two functions as they slide post each other-in other words, in any scanning
operation such as in optical or magnetic recording or picture scanning in television.
Transform theory states that such scanning is equivalent to filtering the signal with a filter
whose transmission is the transform of the scanning function (reversed in time).
Conversely, the effect of an electrical filter is equivalent to a convolution of the input with
a time function that is the transform of filter characteristic. This function, the so-called
"memory curve" of the filter, is identical with the filter impulse response, aside from
dimensions. (Note: the convolution of a time function with a unit impulse gives the some
function times the dimensions of the impulse.)

25
 Properties of Transforms

Time operation Frequency operation Significance

Multiplication Convolution The spectrum of the product of two time functions is the convolution of their spectra.
∞ This is the more general statement of the modulation property. For example, sampling a
f ( t ) g(t ) 1 signal is equivalent to multiplying it by a regular train of unit area impulses. The spectrum
2π −∫∞
F (s)g(ω − s)ds
of the sampled signal consists of the original signal spectrum repeated about each
component of the (line) spectrum of the train of impulses (see pair 6S, page 24). For
no overlap, the highest frequency in the signal to be sampled must be less than half the
sampling frequency. If this is true, original signal spectrum (hence signal) can be recov-
ered with a low pass filter (sampling theorem).

Differentiation Multiplication The spectrum of the nth derivative of a function is (iω)n times the spectrum of the function.
d nf (t ) by p A “differentiating network” has (over the appropriate frequency range) a transmission
p
dt n pnF ( p) K where K is dimensionless or has the dimensions of impedance or admittance.
ω0
Thus, the output wave is proportional to the derivative of the input.

Integration Multiplication The spectrum of the nth integral of a function is (iω)−n times the spectrum of the
t by 1 function. Thus, the response of any filter to a step function is the integral of its impulse
∫∫∫ −∞ f (τ ) (dτ )n p response. An “integrating network” has (over the appropriate frequency range) a trans-
 1 ω0
n F ( p) mission K , where K is dimensionless or has the dimensions of impedance or
pn p
admittance. Thus, the output is proportional to the integral of the past of the input.

26
 Appendix B. IF Amplifier Response and Distortion

IF amplifier response
Mention was made in the test of the phenomenon of decreased
sensitivity and resolution that results when a CW signal is swept
by the IF amplifier at a high rate compared to the bandwidth
squared. Assuming a Gaussian response for the amplifier, the
resulting transient can be determined as follows:

A sweep frequency signal as illustrated in Figure B-1 can be Substituting back for τ and simplifying
represented by  
 T δ2 
F 1− j s
( t ) exp(jπ Ts t 2 )
s= (B-1)
=y (t )
1  1
exp  −
2πFs 
δ 2 t2 
s 2
2πFs  2 T δ2  
1− j 2  1+  s  
Tsδ  2 πF 
using pair 10 of Appendix A   s  (B-6)
 1 
S (ω ) =τ 2π exp  − (τω )2  The envelope of y(t ) is then
 2   
 
jTs 1  1 δ 2 t2 
τ=
where (B-2) =y (t ) exp  − 
2πFs 2  2 T δ2   2
 2πFs  4 s
1+   1+   
2  

 Tsδ    2πFs   (B-7)
If we assume a Gaussian response,
 1  ω 2  Note that for low sweep rates
H (=
ω ) exp  −   
 2  δ   (B-3) Ts 1
   2
the product of S (ω ) H (ω ) gives
2πFs δ

  2
 1 1   2πF  2
Y (ω ) =S (ω ) H (ω ) =τ 2π exp  − τ 2 + 2  ω 2 
2  δ   (B-4) ( t ) exp − 21  T δs
y=  t 

  s   (B-8)
The output transient is the inverse transform of this function,
again using pair 10 This, as was stated earlier, is a plot of the frequency response of
the IF amplifier.
  
 1 τ2
τ 
=y (t ) exp  −  
1  2 τ 2 + 1 
τ2 + 2    (B-5)
δ δ2  

F sweep width
slope = s =
Ts sweep time

Figure B-1. A sweep frequency signal

27
 Distortion In a like manner, the 3 dB bandwidth of the function (B-7) is
2
If the condition on (B-8) is not satisfied, the resulting transient 2 ln(2)  δT 
will be altered in both width (time duration) and amplitude. The =

∆t' 1+  s 
δπ  2πF (B-12)
 s 
reduction in amplitude will be
1
α= The ratio of these times is
2
4  2πFs 
1+  (B-9) 2
 T δ 2  ∆t'  2πFs 
 s  = 1+ 
∆t  T δ 2  (B-13)
π  s 
Noting that δ = B where B is the 3 dB bandwidth,
ln(2)
1 This is the ratio of the effective resolving bandwidth of a spec-
α= trum analyzer to the bandwidth of the IF amplifier as a function of
2 2
4  2ln(2)   Fs  sweep rate. Rewritten in terms of 3 dB bandwidth B.
1+    2 (B-10)
 π   Ts B  2 2
Beff  2ln(2)   Fs 
= 1+    2 
A plot of this function in dB versus – Fs/(Ts B2) is included as B  π   Ts B  (B-14)
Figure B-2.

If we solve for the 3 dB time duration Δt from equation (B-8) by This function is plotted in Figure B-2.
setting α to 1/√2 and solving for the appropriate Δt, we get
Fs = Sweep width
ln(2)δ Ts Ts = Sweep time
∆t = B = 3 dB IF bandwidth
πFs (B-11)
Beff = Effective bandwidth

1000
Loss in amplitude and sensitivity a (dB)

Beff
B

100
Normalized effective bandwidth

5
a
10

20 10

25

30 Beff
B
35
1
0.1 1.0 10 100 1000
Fs
Normalized sweep rate
Ts B2

Figure B-2. Sensitivity loss and normalized effective bandwidth vs. normalized sweep rate

28

Agilent Email Updates
www.agilent.com/find/emailupdates
Get the latest information on the
products and applications you select.
www.lxistandard.org
LAN eXtensions for Instruments puts
the power of Ethernet and the Web
inside your test systems. Agilent
is a founding member of the LXI
consortium.
Agilent Channel Partners
www.agilent.com/find/channelpartners
Get the best of both worlds: Agilent’s
measurement expertise and product
breadth, combined with channel
partner convenience.
Agilent Advantage Services is committed
to your success throughout your equip-
ment’s lifetime. To keep you competitive,
we continually invest in tools and
processes that speed up calibration and
repair and reduce your cost of ownership.
You can also use Infoline Web Services
to manage equipment and services more
effectively. By sharing our measurement
and service expertise, we help you create
the products that change our world.
www.agilent.com/find/advantageservices
www.agilent.com/quality
www.agilent.com
www.agilent.com/find/sa
www.agilent.com/find/X-Series
www.agilent.com/find/psa
www.agilent.com/find/esa
For more information on Agilent
Technologies’ products, applications or
services, please contact your local Agilent
office. The complete list is available at:
www.agilent.com/find/contactus
Americas
Canada (877) 894 4414
Brazil (11) 4197 3600
Mexico 01800 5064 800
United States (800) 829 4444
Asia Pacific
Australia 1 800 629 485
China 800 810 0189
Hong Kong 800 938 693
India 1 800 112 929
Japan 0120 (421) 345
Korea 080 769 0800
Malaysia 1 800 888 848
Singapore 1 800 375 8100
Taiwan 0800 047 866
Other AP Countries (65) 375 8100
Europe & Middle East
Belgium 32 (0) 2 404 93 40
Denmark 45 45 80 12 15
Finland 358 (0) 10 855 2100
France 0825 010 700*
*0.125 €/minute
Germany 49 (0) 7031 464 6333
Ireland 1890 924 204
Israel 972-3-9288-504/544
Italy 39 02 92 60 8484
Netherlands 31 (0) 20 547 2111
Spain 34 (91) 631 3300
Sweden 0200-88 22 55
United Kingdom 44 (0) 118 927 6201
For other unlisted countries:
www.agilent.com/find/contactus
Revised: January 6, 2012
Product specifications and descriptions
in this document subject to change
without notice.
© Agilent Technologies, Inc. 2012
Published in USA, July 5, 2012
5952-1039